GUI complexity

[Bruno Postle]

On Thu 24-Apr-2003 at 03:18:57PM +0200, Pablo d'Angelo wrote:
> On Thu, 24 Apr 2003, Bruno Postle wrote:
> 
> > (With architectural photography, you often want to optimise
> > roll, pitch & yaw for every image)
> 
> but in the end, a global optimization step might change them back
> to something that make them fit together, after you have optimized
> each one of them to get straight skyscrapers, or?

Yep, us architects like to keep all our verticals parallel; in this
picture, only the yaw was fixed for one anchor image:

   http://bugbear.blackfish.org.uk/~bruno/photos/istanbul/blue-mosque-courtyard.jpg

This is not a very good example either, but it's a "panorama" that
consists of one source image where _everything_ has been optimised
(roll, pitch & yaw):

    http://bugbear.blackfish.org.uk/~bruno/photos/dendera/

> > Bear in mind that PTOptimizer optimises images _into_ the output
> > format, so you have to select output parameters before you can
> > optimize.
> 
> Hmm, so the optimization strategy is different for each output
> format?

It can be different, the horizontal and vertical control points
will produce different results with different output formats.

Fisheye images have no straight lines whatsoever; so if your output
is a fisheye, the H & V control points are useless.

Spherical and cylindrical panoramas have no horizontal straight
lines (other than the actual horizon if the photos were taken at
sea), but may have lots of vertical straight lines.

A clever GUI might say:

   "Warning, you have selected horizontal control points; however
   your output panorama is spherical.  Are you sure?"

> I thought it would be the same. Why would it affect the
> optimization, the optimizer "just" determines the position of the
> images and distortion variables. The position of the images do not
> change when selecting a different output format, just the
> reprojection is changed, or?

When stitching/rendering only the reprojection is changed, but these
things make a difference when optimising.

> > Also, once all the input image parameters are decided, this
> > output format is irrelevant, you could "render" into any number
> > of different output formats.
> 
> so why should the output format influence the estimation of image
> parameters (when you say that they are output format independant).

I think that originally it made a lot of difference, the optimizer
based all it's results on the pixel distances in the output image.

I seem to remember that the optimizer later started judging on the
basis of absolute spatial angle (I might be wrong about this).

> On the code side: I'll take a look at wxwin in the next days and
> I'll try to port my control point picker to wxwin, so that I can
> give a more qualified statement than: I just like qt more. After
> that I'll decide if I'll stick to qt or join the wxwindows
> development.

It's easy for me to say (since I'm not offerring to write any C++),
but it might be cool to have two GUI tools ;-)

-- 
Bruno